info

Converted a simple model from keras which gets N images an input and outputs M images. From Android app (java) I'm loading images as bitmaps and converting the bitmaps into TensorImages, which are later joined together into a TensorImage array.

case 1

Trying to input the TensorImage array to the model using tflite.run(inputArray, outputArray) when inputArray is TensorImage inputArray[] = new TensorImage[]{tensorImage1, tensorImage2, tensorImage3} results in the following DataType error: DataType error: cannot resolve DataType of org.tensorflow.lite.support.image.TensorImage, Which I assume has to do with the input buffer limitations.

case 2

Trying to input a single TensorImage's buffer using tflite.run(tensorImage1.getBuffer(), output.getBuffer()) results in an Illegal Argument Exception: IllegalArgumentException: Both buffer and bitmap data are obsolete.

case 3

Changed the output to a TensorBuffer and created a float32 buffer like so: TensorBuffer result = TensorBuffer.createFixedSize(new int[]{1, 1, 10}, DataType.FLOAT32) which works incorrectly since the model is applied on a single image (though it provides an output).

question

So how should images be given as input into a tflite model?